Paul Durrant writes:
> a) Mapping a dbllk for DMA and thereby holding onto it, and hence
> holding onto the cred_t.
> b) Passing up packets in desballoc()ed buffers.

Note that both of these contain issues that go beyond just Zones.  The
cred_t holding issue is relatively minor once you get past Zones (the
only other thing the cred holds down are TX labels), but the desballoc
issue can be more significant due to DR.  You can't unload the driver
until you've somehow scavenged away those pending callbacks, and if
there are buffers rotting in someone's input queue (think: SIGSTOP
sent to snoop), you can end up either blocking the operation "forever"
or, if you miscount, getting a callback into freed memory.  Both are
unpleasant.

I don't know if anyone is working on that problem, but we'll likely
have to do "something" about it, as others have hit it.

-- 
James Carlson, KISS Network                    <[EMAIL PROTECTED]>
Sun Microsystems / 1 Network Drive         71.232W   Vox +1 781 442 2084
MS UBUR02-212 / Burlington MA 01803-2757   42.496N   Fax +1 781 442 1677
_______________________________________________
networking-discuss mailing list
[email protected]

Reply via email to